IsPixelWorking for AI agents

Connect your AI assistant to IsPixelWorking and ask it to check a page's Meta Pixel, explain the report and walk you through the fix. It works with any assistant that supports remote MCP servers: Claude, ChatGPT, Cursor, VS Code and others.

Server address

https://ispixelworking.com/mcp

A remote MCP server (Streamable HTTP). No sign-in is needed for checks; an API key from your account raises the limits and adds your monitored pages.

Connect

  • Claude Code: claude mcp add --transport http ispixelworking https://ispixelworking.com/mcp
  • Claude (claude.ai and the desktop app): Settings → Connectors → Add custom connector, with the address above.
  • ChatGPT: add a custom connector (developer mode) with the address above.
  • Cursor: in ~/.cursor/mcp.json, add "ispixelworking": { "url": "https://ispixelworking.com/mcp" } under mcpServers.
  • VS Code: in .vscode/mcp.json, add "ispixelworking": { "type": "http", "url": "https://ispixelworking.com/mcp" } under servers.

What to ask

  • "Check our landing page with IsPixelWorking and explain what's wrong." (give it the address)
  • "My Meta Pixel fires before cookie consent on Framer. Get the fix guide and walk me through it."
  • "I published the fix. Run the check again."
  • "Which of my monitored pages have open incidents?" (with an API key on a monitoring plan)

Tools

  • check_page: checks a page and waits for the result, usually 20 to 60 seconds. A result from the last 24 hours is reused; a fresh check is allowed once the last one is an hour old.
  • get_report: reads a report by its id or link: grade, problems with explanations and evidence, guide links, checklist.
  • get_fix_guide: the step-by-step guide for a problem, for your platform (Framer, Shopify, Webflow, WordPress) or your consent tool (Cookiebot, CookieYes, OneTrust, Usercentrics, Complianz).
  • list_findings: every problem the check can report.
  • With an API key on a monitoring plan: list_monitored_pages, recheck_monitored_page and add_monitored_page.

The check is the same as on the website: a browser in the EU visits the page twice, never sends tracking data to Meta, TikTok or Google, never fills in forms and never buys anything. See how the check works.

Limits

  • Without a key: 5 new checks a day from one IP address, the same as the website, and a shared daily cap for all agents. Results from the last 24 hours are free.
  • With a key: your plan's checks per day: Free 5, Pro 20, Agency 100. Pro and Agency keys also get the monitoring tools.

API keys

Create a key in your account under Alerts and domains → API keys for AI agents. The key is shown once; anyone who has it can use your limits, so keep it private and revoke it there if it leaks.

Send it as a header, Authorization: Bearer ipw_…. In Claude Code:

claude mcp add --transport http ispixelworking https://ispixelworking.com/mcp --header "Authorization: Bearer ipw_…"

In Cursor and VS Code, add "headers": { "Authorization": "Bearer ipw_…" } next to the url. Assistants that only take an address (claude.ai, ChatGPT) connect without a key.

For developers

The server publishes an MCP Server Card and is listed in /.well-known/ai-catalog.json and /.well-known/mcp.json. The agent guide and llms.txt describe the rest of the site; every page also has a Markdown version.